There are many ways to make tpa6120 stable at low overall output impedance.Hope it helps, but Russ White got an 10mH inductor in parallel with a 22Ohms resistor: http://www.twistedpearaudio.com/docs/linestages/ventus_ez_schematic.pdf, others found a different way to reduce output impedance without affecting the stability.
PA probably did something similar, otherwise couldn't achieve 3mOhms of output impedance: http://sjostromaudio.com/pages/index.php/hifi-projects/128-qrv07-headphone-amp?start=6.
L.E.: Actually I don't know how PA achieved that low output imepdance, maybe he simply connected the TPA6120A directly to the headphones jack after all: http://sjostromaudio.com/pages/index.php/hifi-projects/128-qrv07-headphone-amp?showall=1 and http://sjostromaudio.com/hifi_pics/hifi_100pr/qrv07r0_TPA6120.jpg might help, but I see that the two 10.2Ohms blue resistors are connected directly to the output pins and I guess from J1/J2 jumpers these resistors could be shorted. Also, on QRV09 the output impedance is 10/120Ohms: http://sjostromaudio.com/pages/index.php/hifi-projects/130-qrv09-headphone-amp?start=5
1, your first way.
2, zobel networks.
3, composite amplifier.
I'm not sure how fiio did it. But i have a tpa6120 amp has 0.0001% distortion from 20hz-2khz(rises afterwards) using composite amplifier topology. And it's very stable.